Is this the future of opera in Scotland?

It Was on the cards that sooner or later James MacMillan would write a full-scale opera - his past work pointed the way with the progressive insistence of countdown markers to an exit on the M1 - and that it would premiere at the Edinburgh Festival (which has repeatedly backed him as the future of Scottish music). It was also likely that it would be steeped in Catholicism (which to MacMillan is as Orthodoxy is to John Tavener). So Ines de Castro, which opened last weekend at the Festival Theatre, was no real surprise: a definitive statement of where, how and for what its composer stands - and that's important to know, because MacMillan's standing is high. He is one of the few "serious" composers of our time to have caught the public imagination without diluting the complexity of his language down to mystical nothingness or inflating it with the Purcellian equivalent of a disco beat. His success has been based on the theatrically effective espousal of big gestures, big emotions, big causes, and on adherence to a style of writing which is contemporary but in touch with the established sound world of tonality where listeners feel comfortable. It doesn't take post-graduate applied mathematics to absorb his structures. He can hold an audience with words of not much more than one syllable, and clearly learnt how to do it through the discipline of writing for the modern, post-Vatican II church, where the journey to transcendence starts on Earth and runs via Mrs Average in the back pew.

All this you find in Ines, which is a powerful response to a powerful story and a strong, wiry libretto, adapted from a play by John Clifford. The two acts climax well - MacMillan is a master of good endings - and the palette of orchestral colour is ingeniously diverse. But still it doesn't make the impact that it should, largely because the piece doesn't cohere as a totality. Act I is too much a curtain-raiser for Act II, which then leaves you waiting for an Act III that doesn't happen - and worse still, wondering why the piece actually needs to be sung (the killer question no opera can afford to raise).

In one sense there's every excuse for singing. The setting is 14th-century Portugal, where life is ruled by the church, its liturgy and its chant - which features prominently in MacMillan's score. He obviously intended the action to be ritualistic and to draw a parallel between the religious icon of Our Lady of Sorrows and the predicament of his heroine Ines, whose misfortune is to be the Spanish mistress of a Portuguese Prince at a time when Spain and Portugal are at war. Learning the lesson of all Other Women in royal marriages, she gets a bad press and is murdered with the approval of the Prince's father (a solution to the Parker Bowles affair uncanvassed, so far, by the British tabloids), after which the Prince takes terrible revenge with one of the more anatomically inventive executions ever devised. The opera ends spectacularly with a Marian coronation of the corpse of Ines and a still more Marian appearance of her spirit to a young girl. At least, it should be spectacular.

But Jonathan Moores's production is prosaic, in a Shakespearian low- life way that gives no true sense of time, place or atmosphere. Chris Dyer's designs look cheaply amateur (it's clear that Scottish Opera hasn't spent much on the show) and very little of the visual presentation could be called convincing. So the music has to carry everything, and on first hearing it just didn't - despite competent conducting (Richard Armstrong) and performances. Helen Field commanded some sympathy as Ines, but made less of an impression than Christopher Purves in the comparatively minor role of her executioner.

If I had to identify the basic problem with the score of Ines, I'd say it looks to the wrong operatic models - the brutality of Shostakovich, earnestness of Hindemith, and beatific stasis of Vaughan Williams - which all have something to offer but not as much as the chilling economy of Britten in The Turn of the Screw, which is more the sort of piece Ines needed to be. And leaving Edinburgh (with relief, because it really wasn't very good this year), there has been an excellent little Turn of the Screw down at Broomhill, the not-so-smart country-house opera venue in Kent. The instrumental playing was unfocused and Charles Hazelwood conducted it like Puccini; but the staging (Caroline Ward) and simple monochrome designs (Jane Singleton) had the hairs of my neck on end, with ghosts that sprang out of the darkness or appeared in silhouette, and an intensity that choked the small stage with accumulative force. Lynne Davies sang a superb Governess; Beverley O'Regan Thiele made a lugubriously fallen Miss Jessel; and Caroline Rowlands almost stole the show as an uncommonly expressive Mrs Grose.

The week's Proms brought a new symphony by the German protege of Henze, Detlev Glanert, which was lighter in weight and more conventional than expected, with five short, fancifully titled movements that didn't quite live up to its promise as an essay on the nature of symphonic writing. But it showed a virtuosic mastery of larger forces - the BBC Scottish SO under Osmo Vanska - and a more engaging liveliness than most new music coming out of Germany.

It's a well-known paradox that the great English oratorios are written by Germans, and whether or not you consider Mendelssohn's Elijah great, it was revered by the Victorians and positively belongs in the Albert Hall - where Richard Hickox on Monday conducted a performance so fine it almost circumvented my impatience with the pseudo-Baroque counterpoint and sanctimonious arias. In fact, Hickox went as far as you can with this piece to de-Victorianise it, ironing out the sentiment and cleaning out the gunge. Despite a massive chorus and no fewer than 10 singers sharing out the solo roles, it was crisp and buoyant, with Bryn Terfel tiring slightly as the prophet but otherwise magnificent. The National Orchestra of Wales, too, was in good form.

But good form for a British out-of-London orchestra is still several existences short of the nirvana of the Berlin Philharmonic, who gave two Proms this week under Claudio Abbado. It's smart nowadays to talk down the legacy of Karajan and the might of his orchestra, but Wednesday's Brahms programme and Thursday's Mahler 2 confirmed that the show really does go on. And this despite some imperfections of ensemble which wouldn't pass the producer's pen in a recording but are, in concert, an almost natural consequence of the flexible, selective and anticipatory nature of Abbado's beat. He gives his players licence; but with musicians of such calibre in every department of the band, from timps to basses, it isn't such a risk. The Brahms 1st Piano Concerto took a while to come together, and Radu Lupu, the side-saddle pianist - half off his seat and oddly disengaged - can't be the easiest collaborator. But the Brahms 1st Symphony was beautifully realised - the whole thing weighed, balanced and held complete like some exquisite object in Abbado's palm - and the Mahler was a masterpiece of prophylactic understatement, holding back until the final moments, which arrived the grander for the contrast. All musical London turned out for these concerts; and if they weren't quite benchmark performances, they were certainly formidable. You don't get many like them.